Anders Eklund

Johan Jönemo, Muhammad Usman Akbar, Robin Kämpe et al.

Using 3D CNNs on high resolution medical volumes is very computationally demanding, especially for large datasets like the UK Biobank which aims to scan 100,000 subjects. Here we demonstrate that using 2D CNNs on a few 2D projections (representing mean and standard deviation across axial, sagittal and coronal slices) of the 3D volumes leads to reasonable test accuracy when predicting the age from brain volumes. Using our approach, one training epoch with 20,324 subjects takes 20 - 50 seconds using a single GPU, which two orders of magnitude faster compared to a small 3D CNN. These results are important for researchers who do not have access to expensive GPU hardware for 3D CNNs.

Måns Larsson, Muhammad Usman Akbar, Anders Eklund

Large annotated datasets are required to train segmentation networks. In medical imaging, it is often difficult, time consuming and expensive to create such datasets, and it may also be difficult to share these datasets with other researchers. Different AI models can today generate very realistic synthetic images, which can potentially be openly shared as they do not belong to specific persons. However, recent work has shown that using synthetic images for training deep networks often leads to worse performance compared to using real images. Here we demonstrate that using synthetic images and annotations from an ensemble of 20 GANs, instead of from a single GAN, increases the Dice score on real test images with 4.7 % to 14.0 % on specific classes.

Shreyas Fadnavis, Jens Sjölund, Anders Eklund et al.

Diffusion MRI (dMRI) is the only non-invasive technique sensitive to tissue micro-architecture, which can, in turn, be used to reconstruct tissue microstructure and white matter pathways. The accuracy of such tasks is hampered by the low signal-to-noise ratio in dMRI. Today, the noise is characterized mainly by visual inspection of residual maps and estimated standard deviation. However, it is hard to estimate the impact of noise on downstream tasks based only on such qualitative assessments. To address this issue, we introduce a novel metric, Noise Uncertainty Quantification (NUQ), for quantitative image quality analysis in the absence of a ground truth reference image. NUQ uses a recent Bayesian formulation of dMRI models to estimate the uncertainty of microstructural measures. Specifically, NUQ uses the maximum mean discrepancy metric to compute a pooled quality score by comparing samples drawn from the posterior distribution of the microstructure measures. We show that NUQ allows a fine-grained analysis of noise, capturing details that are visually imperceptible. We perform qualitative and quantitative comparisons on real datasets, showing that NUQ generates consistent scores across different denoisers and acquisitions. Lastly, by using NUQ on a cohort of schizophrenics and controls, we quantify the substantial impact of denoising on group differences.

Muhammad Usman Akbar, Måns Larsson, Anders Eklund

Large annotated datasets are required for training deep learning models, but in medical imaging data sharing is often complicated due to ethics, anonymization and data protection legislation. Generative AI models, such as generative adversarial networks (GANs) and diffusion models, can today produce very realistic synthetic images, and can potentially facilitate data sharing. However, in order to share synthetic medical images it must first be demonstrated that they can be used for training different networks with acceptable performance. Here, we therefore comprehensively evaluate four GANs (progressive GAN, StyleGAN 1-3) and a diffusion model for the task of brain tumor segmentation (using two segmentation networks, U-Net and a Swin transformer). Our results show that segmentation networks trained on synthetic images reach Dice scores that are 80% - 90% of Dice scores when training with real images, but that memorization of the training images can be a problem for diffusion models if the original dataset is too small. Our conclusion is that sharing synthetic medical images is a viable option to sharing real images, but that further work is required. The trained generative models and the generated synthetic images are shared on AIDA data hub

Marco Domenico Cirillo, David Abramian, Anders Eklund

Training segmentation networks requires large annotated datasets, which in medical imaging can be hard to obtain. Despite this fact, data augmentation has in our opinion not been fully explored for brain tumor segmentation. In this project we apply different types of data augmentation (flipping, rotation, scaling, brightness adjustment, elastic deformation) when training a standard 3D U-Net, and demonstrate that augmentation significantly improves the network's performance in many cases. Our conclusion is that brightness augmentation and elastic deformation work best, and that combinations of different augmentation techniques do not provide further improvement compared to only using one augmentation technique. Our code is available at https://github.com/mdciri/3D-augmentation-techniques.

Per Welander, Simon Karlsson, Anders Eklund

In medical imaging, a general problem is that it is costly and time consuming to collect high quality data from healthy and diseased subjects. Generative adversarial networks (GANs) is a deep learning method that has been developed for synthesizing data. GANs can thereby be used to generate more realistic training data, to improve classification performance of machine learning algorithms. Another application of GANs is image-to-image translations, e.g. generating magnetic resonance (MR) images from computed tomography (CT) images, which can be used to obtain multimodal datasets from a single modality. Here, we evaluate two unsupervised GAN models (CycleGAN and UNIT) for image-to-image translation of T1- and T2-weighted MR images, by comparing generated synthetic MR images to ground truth images. We also evaluate two supervised models; a modification of CycleGAN and a pure generator model. A small perceptual study was also performed to evaluate how visually realistic the synthesized images are. It is shown that the implemented GAN models can synthesize visually realistic MR images (incorrectly labeled as real by a human). It is also shown that models producing more visually realistic synthetic images not necessarily have better quantitative error measurements, when compared to ground truth data. Code is available at https://github.com/simontomaskarlsson/GAN-MRI

Anders Eklund

Training of multimodal foundation models is currently restricted to centralized data centers containing massive, aligned datasets (e.g., image-text pairs). However, in realistic federated environments, data is often unpaired and fragmented across disjoint nodes; one node may hold sensor data, while another holds textual logs. These datasets are strictly private and share no common samples. Current federated learning (FL) methods fail in this regime, as they assume local clients possess aligned pairs or require sharing raw feature embeddings, which violates data sovereignty. We propose a novel framework to train a global multimodal transformer across decentralized nodes with disjoint modalities. We introduce a small public anchor set to align disjoint private manifolds. Using Gram matrices calculated from these public anchors, we enforce semantic alignment across modalities through centered kernel alignment without ever transmitting private samples, offering a mathematically superior privacy guarantee compared to prototype sharing. Further, we introduce a subspace-stabilized fine-tuning method to handle FL with huge transformer models. We strictly decouple domain-specific magnitude shifts from semantic direction, ensuring that nodes with varying sensor characteristics align geometrically to the global consensus. Lastly, we propose precision weighted averaging, where efficiently obtained uncertainty estimates are used to downweight uncertain nodes. This paper establishes the mathematical backbone for federated unpaired foundation models, enabling a global model to learn a unified representation of the world from fragmented, disjoint, and private data silos without requiring centralized storage or paired samples.

Muhammad Usman Akbar, Wuhao Wang, Anders Eklund

Diffusion models were initially developed for text-to-image generation and are now being utilized to generate high quality synthetic images. Preceded by GANs, diffusion models have shown impressive results using various evaluation metrics. However, commonly used metrics such as FID and IS are not suitable for determining whether diffusion models are simply reproducing the training images. Here we train StyleGAN and a diffusion model, using BRATS20, BRATS21 and a chest x-ray pneumonia dataset, to synthesize brain MRI and chest x-ray images, and measure the correlation between the synthetic images and all training images. Our results show that diffusion models are more likely to memorize the training images, compared to StyleGAN, especially for small datasets and when using 2D slices from 3D volumes. Researchers should be careful when using diffusion models (and to some extent GANs) for medical imaging, if the final goal is to share the synthetic images.

Iulian Emil Tampu, Anders Eklund, Neda Haj-Hosseini

In the application of deep learning on optical coherence tomography (OCT) data, it is common to train classification networks using 2D images originating from volumetric data. Given the micrometer resolution of OCT systems, consecutive images are often very similar in both visible structures and noise. Thus, an inappropriate data split can result in overlap between the training and testing sets, with a large portion of the literature overlooking this aspect. In this study, the effect of improper dataset splitting on model evaluation is demonstrated for three classification tasks using three OCT open-access datasets extensively used, Kermany's and Srinivasan's ophthalmology datasets, and AIIMS breast tissue dataset. Results show that the classification performance is inflated by 0.07 up to 0.43 in terms of Matthews Correlation Coefficient (accuracy: 5% to 30%) for models tested on datasets with improper splitting, highlighting the considerable effect of dataset handling on model evaluation. This study intends to raise awareness on the importance of dataset splitting given the increased research interest in implementing deep learning on OCT data.

Johan Jönemo, David Abramian, Anders Eklund

Classifying subjects as healthy or diseased using neuroimaging data has gained a lot of attention during the last 10 years. Here we apply deep learning to derivatives from resting state fMRI data, and investigate how different 3D augmentation techniques affect the test accuracy. Specifically, we use resting state derivatives from 1,112 subjects in ABIDE preprocessed to train a 3D convolutional neural network (CNN) to perform the classification. Our results show that augmentation only provide minor improvements to the test accuracy.

Iulian Emil Tampu, Neda Haj-Hosseini, Anders Eklund

Effective, robust, and automatic tools for brain tumor segmentation are needed for the extraction of information useful in treatment planning from magnetic resonance (MR) images. Context-aware artificial intelligence is an emerging concept for the development of deep learning applications for computer-aided medical image analysis. In this work, it is investigated whether the addition of contextual information from the brain anatomy in the form of white matter, gray matter, and cerebrospinal fluid masks and probability maps improves U-Net-based brain tumor segmentation. The BraTS2020 dataset was used to train and test two standard 3D U-Net models that, in addition to the conventional MR image modalities, used the anatomical contextual information as extra channels in the form of binary masks (CIM) or probability maps (CIP). A baseline model (BLM) that only used the conventional MR image modalities was also trained. The impact of adding contextual information was investigated in terms of overall segmentation accuracy, model training time, domain generalization, and compensation for fewer MR modalities available for each subject. Results show that there is no statistically significant difference when comparing Dice scores between the baseline model and the contextual information models, even when comparing performances for high- and low-grade tumors independently. Only in the case of compensation for fewer MR modalities available for each subject did the addition of anatomical contextual information significantly improve the segmentation of the whole tumor. Overall, there is no overall significant improvement in segmentation performance when using anatomical contextual information in the form of either binary masks or probability maps as extra channels.

Mehdi Foroozandeh, Anders Eklund

Training segmentation networks requires large annotated datasets, but manual annotation is time consuming and costly. We here investigate if the combination of a noise-to-image GAN and an image-to-image GAN can be used to synthesize realistic brain tumor images as well as the corresponding tumor annotations (labels), to substantially increase the number of training images. The noise-to-image GAN is used to synthesize new label images, while the image-to-image GAN generates the corresponding MR image from the label image. Our results indicate that the two GANs can synthesize label images and MR images that look realistic, and that adding synthetic images improves the segmentation performance, although the effect is small.

Marco Domenico Cirillo, David Abramian, Anders Eklund

Gliomas are the most common primary brain malignancies, with different degrees of aggressiveness, variable prognosis and various heterogeneous histological sub-regions, i.e., peritumoral edema, necrotic core, enhancing and non-enhancing tumour core. Although brain tumours can easily be detected using multi-modal MRI, accurate tumor segmentation is a challenging task. Hence, using the data provided by the BraTS Challenge 2020, we propose a 3D volume-to-volume Generative Adversarial Network for segmentation of brain tumours. The model, called Vox2Vox, generates realistic segmentation outputs from multi-channel 3D MR images, segmenting the whole, core and enhancing tumor with mean values of 87.20%, 81.14%, and 78.67% as dice scores and 6.44mm, 24.36mm, and 18.95mm for Hausdorff distance 95 percentile for the BraTS testing set after ensembling 10 Vox2Vox models obtained with a 10-fold cross-validation.

Anders Eklund

Deep learning requires large datasets for training (convolutional) networks with millions of parameters. In neuroimaging, there are few open datasets with more than 100 subjects, which makes it difficult to, for example, train a classifier to discriminate controls from diseased persons. Generative adversarial networks (GANs) can be used to synthesize data, but virtually all research is focused on 2D images. In medical imaging, and especially in neuroimaging, most datasets are 3D or 4D. Here we therefore present preliminary results showing that a 3D progressive growing GAN can be used to synthesize MR brain volumes.

David Abramian, Anders Eklund

Registration between an fMRI volume and a T1-weighted volume is challenging, since fMRI volumes contain geometric distortions. Here we present preliminary results showing that 3D CycleGAN can be used to synthesize fMRI volumes from T1-weighted volumes, and vice versa, which can facilitate registration.

Sevil Maghsadhagh, Anders Eklund, Hamid Behjat

The human brain cortical layer has a convoluted morphology that is unique to each individual. Characterization of the cortical morphology is necessary in longitudinal studies of structural brain change, as well as in discriminating individuals in health and disease. A method for encoding the cortical morphology in the form of a graph is presented. The design of graphs that encode the global cerebral hemisphere cortices as well as localized cortical regions is proposed. Spectral metrics derived from these graphs are then studied and proposed as descriptors of cortical morphology. As proof-of-concept of their applicability in characterizing cortical morphology, the metrics are studied in the context of hemispheric asymmetry as well as gender dependent discrimination of cortical morphology.

Sevil Maghsadhagh, Mousa Shamsi, Anders Eklund et al.

The human brain cortical layer has a convoluted morphology that is unique to each individual. Characterization of the cortical morphology is necessary in longitudinal studies of structural brain change, as well as in discriminating individuals in health and disease. A method for encoding the cortical morphology in the form of a graph is presented. The design of graphs that encode the global cerebral hemisphere cortices as well as localized cortical regions is proposed. Spectral features of these graphs are then studied and proposed as descriptors of cortical morphology. As proof-of-concept of their applicability in characterizing cortical morphology, the descriptors are studied in the context of discriminating individuals based on their sex.

David Abramian, Anders Eklund

Anonymization of medical images is necessary for protecting the identity of the test subjects, and is therefore an essential step in data sharing. However, recent developments in deep learning may raise the bar on the amount of distortion that needs to be applied to guarantee anonymity. To test such possibilities, we have applied the novel CycleGAN unsupervised image-to-image translation framework on sagittal slices of T1 MR images, in order to reconstruct facial features from anonymized data. We applied the CycleGAN framework on both face-blurred and face-removed images. Our results show that face blurring may not provide adequate protection against malicious attempts at identifying the subjects, while face removal provides more robust anonymization, but is still partially reversible.

Xuan Gu, Hans Knutsson, Markus Nilsson et al.

Diffusion magnetic resonance imaging (diffusion MRI) is a non-invasive microstructure assessment technique. Scalar measures, such as FA (fractional anisotropy) and MD (mean diffusivity), quantifying micro-structural tissue properties can be obtained using diffusion models and data processing pipelines. However, it is costly and time consuming to collect high quality diffusion data. Here, we therefore demonstrate how Generative Adversarial Networks (GANs) can be used to generate synthetic diffusion scalar measures from structural T1-weighted images in a single optimized step. Specifically, we train the popular CycleGAN model to learn to map a T1 image to FA or MD, and vice versa. As an application, we show that synthetic FA images can be used as a target for non-linear registration, to correct for geometric distortions common in diffusion MRI.

Jens Sjölund, Anders Eklund, Evren Özarslan et al.

We propose to use Gaussian process regression to accurately estimate the diffusion MRI signal at arbitrary locations in q-space. By estimating the signal on a grid, we can do synthetic diffusion spectrum imaging: reconstructing the ensemble averaged propagator (EAP) by an inverse Fourier transform. We also propose an alternative reconstruction method guaranteeing a nonnegative EAP that integrates to unity. The reconstruction is validated on data simulated from two Gaussians at various crossing angles. Moreover, we demonstrate on non-uniformly sampled in vivo data that the method is far superior to linear interpolation, and allows a drastic undersampling of the data with only a minor loss of accuracy. We envision the method as a potential replacement for standard diffusion spectrum imaging, in particular when acquistion time is limited.